Battle of Ideas 2011, 29th October 2011, Royal College of Art, London
journalist and writer; co-writer and performer, Edinburgh Festival Fringe smash hit Your Days Are Numbered: the maths of death
actuarial partner, London, PwC
author, Velvet Glove, Iron Fist and The Spirit Level Delusion: fact-checking the Left's new theory of everything
founder, First Actuarial plc; chair, Manchester Salon
Warning: may contain graphic images of data.
From Hans Roslingâs surprise hit TV show The Joy of Stats to visual snapshots of the numbers in the news, there seems to be a new love affair with statistics, especially when they come in a graphic form that can be grasped intuitively. It doesnât take a maths degree to see that a straight line sloping upwards suggests a relationship between the height and age of schoolchildren, for example, or that a large blob represents more murders than a smaller blob. We are ready, perhaps too ready, to give credence to statistics, which appear to manifest the mysterious labours of the mathematically literate in transparent, self-explanatory form. We too rarely question the assumptions that underlie the figures, and too often forget that an evident correlation between two things â US oil production and the quality of rock ânâ roll, for example â does not necessarily mean thereâs a causal relationship. And we are often beguiled into believing that the past automatically predicts the future â a graph which shows anything increasing exponentially in the past can only spell doom for the future.
We need statistics. By collecting lots of simple information in numerical form we can see patterns that may help us understand problems and spot underlying causes. But this is where things get tricky. If the number of Elvis Presley impersonators continues to increase as it did from 1957 (170 worldwide) to 2007 (over 85,000 worldwide) one in three of us will be Elvis impersonators by 2019. Is this likely? Public health and economics, among other disciplines, rely on modelling human behaviour the same way animal behaviour, or the behaviour of water molecules, can be modelled: by looking at what theyâve done so far. Not surprisingly, this leaves the predicted future looking very much like the past. But people are not data points, and both individuals and societies can behave in unpredictable ways. You can calculate your probability of living to be 100 (one in six of the current UK population) but thatâs an educated estimate of the odds, not a guarantee. Are we in danger of turning statistical modelling from a useful analytical tool to the new astrology?